As enterprises connect AI agents to more operational systems and sensitive data, determining what those systems can access — and under whose permissions — is becoming an important part of moving AI into production.
A recent CData article, What to Look for in a Context Layer, explores the infrastructure needed to give AI applications accurate, current, and permission-aware access to enterprise data. The article evaluates context-layer architecture across source coverage, schema intelligence, permission enforcement, and measurable accuracy.
Why Permission Enforcement Matters for Enterprise AI
CData highlights permission enforcement as a key consideration when AI applications and agents interact with enterprise systems. Rather than giving every agent broad access through shared credentials, the article argues that data access should reflect the permissions of the individual making the request.
The challenge is finding the right balance. Giving AI systems too much access can expose information that should remain restricted. Limiting access too aggressively can prevent AI from using the enterprise data and context required to deliver useful results.
Protegrity Research Highlights the Cost of Over-Restricting AI
To illustrate that challenge, CData cites research commissioned by Protegrity and conducted by Enterprise Management Associates (EMA). The research found that 81.6% of organizations reported deploying AI in a diminished state because of security concerns, including reduced agent autonomy or restricted access to critical enterprise data.
The finding is part of Protegrity’s The State of AI Friction: Why Enterprise AI Deployment Is Slower, Costlier, and More Limited Than Expected research, which examines how security, governance, and compliance requirements can contribute to delayed deployments, reduced AI functionality, and additional operational overhead.
Building Governance Into the Data Layer
The CData article reinforces a broader challenge facing enterprise AI teams: governance needs to work where AI actually accesses and uses data. As agents interact with CRM, ERP, HR, analytics, and other operational systems, organizations need a clear understanding of what information is available, who or what is authorized to use it, and how those policies are applied.
For enterprises moving AI from pilot to production, the goal is not simply to provide more access or impose more restrictions. It is to create governed access that allows AI to use the data it needs while maintaining control over sensitive information.
Note: This summary is based on the external CData article “What to Look for in a Context Layer” and is provided for convenience. Please refer to the original publication for full context and source reporting.